Juicedata

JuiceFS, an #opensource #cloudnative #distributed #filesystem, compatible with POSIX, HDFS and S3 API.

[Juicedata @JuicedataInc]

Just published: a @CloudNativeFdn-featured case study on how @StepFun_ai scaled multimodal AI storage beyond 10,000 JuiceFS clients with Kubernetes and TiKV. Proud that production-driven optimizations landed upstream in JuiceFS.

#KubeCon #CloudNativeCon

https://t.co/kJlSyG5UcQ

[Juicedata @JuicedataInc]

Catch our #KubeCon keynote on the main stage tomorrow morning!

[Franky. @FrankYouChill]

In Salt Lake City for @KubeCon_ this year! Really wanna see what’s next for AI agent infra.

I will speaking with Joe Zhou from @juicedatainc on the keynote stage on November 12th. We will explain how @StepFun_ai scaled a shared storage layer for multimodal training and inference to 10,000+ hosts.

If you’re also in KubeCon, DM me and let’s have a chat!!

[Juicedata @JuicedataInc]

Exactly. Agent workloads need more than object storage: efficient metadata operations, POSIX access, and cache reuse.

We recently wrote about our experience and ongoing optimization work:

Scaling Data Management for Agent Sandboxes with JuiceFS https://t.co/WESft5C0Nb

[Ashish Tyagi @Ashish_16_2000]

@markokraemer Stateless compute makes sense but agents spam tiny file ops (git status, uv sync) and object storage hates that. You'd need a metadata layer like JuiceFS or a local cache of the hot tree, no?

[Juicedata @JuicedataInc]

👀 What does it take to handle EB-scale data, hundreds of billions of files, and hundreds of thousands of clients?

🚘 See how Horizon Robotics uses #JuiceFS to address metadata bottlenecks, large-scale deletion, training performance, and #MultiTenantAccess.
https://t.co/RsT74c9YJa

[Juicedata @JuicedataInc]

@jerrychaox 我们只支持传统macfuse内核拓展，fuset或者fskit现在都不支持，有一些限制。可以到github讨论区或者issue问

[Rui Su @JuiceFS @suavesu]

Huge congrats to @fal on the new funding round! 🚀

I still remember visiting @isidentical in San Francisco two years ago, when fal was just a tiny team of 4–5 people.

They’ve been riding a rocket ship ever since — and it looks like they’re just getting started! 🔥

[Juicedata @JuicedataInc]

Have experience with JuiceFS? These AI infrastructure teams are hiring 👇
@fal
https://t.co/cVSk9onyCK

@baseten https://t.co/OybtnAcd7O

@FireworksAI_HQ https://t.co/lpYLXBcHVM

Check out the roles—or share with someone who’d be a great fit!

[Juicedata @JuicedataInc]

🚗 #AutonomousDriving R&D needs more than storage. Compute needs data, ready on time.

👇 See how Zhuoyu, an #IntelligentDriving tech company, uses #JuiceFS to manage hundreds of petabytes with #DistributedCaching, proactive warm-up, and optimized reads.

https://t.co/PVypF1EjaV

[Juicedata @JuicedataInc]

JuiceFS Caching: Design Principles, Performance, and Flexibility

Like most things in engineering, caching is all about tradeoffs. This video walks through the design principles behind JuiceFS' flexible caching mechanism and shows how to tune for performance, consistency, and other tradeoffs for your workload.

#Cache #Performance
#DataConsistency #DistributedStorage

[Juicedata @JuicedataInc]

JuiceFS Mirror File System in 3 minutes:
✅ One source region, multiple mirrors
🔁 Data stays in sync across clouds
☁️ Multi-cloud AI, data at your side

[Juicedata @JuicedataInc]

#AIAgent sandboxes are short-lived. Their data isn't.
👇 From #CrossCloud access and isolated mounts to cache reuse and client scaling, discover how #JuiceFS helps manage data across #AgentSandboxes—and the challenges that emerge at scale. #CloudNative
https://t.co/56ZICTaOxe

[Juicedata @JuicedataInc]

Office Hours #15 – JuiceFS Mirror File System: One Volume, Any Cloud https://t.co/seyXC9KZwr

[Juicedata @JuicedataInc]

👉 https://t.co/68j01deS4Z

Join the full 2-day online event at P99 CONF starting tomorrow! JuiceFS has an instant access session, along with tons of great talks on eBPF, Rust, Go, databases, AI, distributed storage, and more.

#ScyllaDB #P99CONF #JuiceFS

[Juicedata @JuicedataInc]

🤔How do #S3Gateway and #WebDAV access the same #JuiceFS #FileSystem? Explore their shared client initialization, protocol-to-filesystem mappings, and how each handles metadata, permissions, and file locks.
https://t.co/p3WUd51nTM

[Juicedata @JuicedataInc]

We are heading to Tech Fuse DSM this week. Looking forward to seeing everyone in Des Moines soon!

[Juicedata @JuicedataInc]

🚀 #AIInference container startup: 116s → 1.4s

#Nydus + #JuiceFS combine on-demand image loading with #DistributedCaching to cut startup data reads from 11 GB to 11 MB. No more registry bottlenecks.

#AIStorage #FileSystem #DistributedStorage

https://t.co/i3wIiY6CCs